gtk-2.2.2 and garbled graphics - Page 3

amthrax · 06-15-03, 03:01 AM

I'm not sure if this will help any, but I was recently debugging the Fluxbox 0.9.3 menu translucency code and discovered that the bug I was tracking was actually in nVidia's render extension. Now that I've disabled it the Fluxbox code works as expected.

The problem appears to be related to the src_x and src_y parameters to XRenderComposite. The destination coordinates behave as expected (I wasn't able to test the mask coordinates from the Fluxbox code). However, the x component of the source coordinate appears to be relative to the right side of the source image, so the way to get it to look right is to use (src_width-src_x). src_y has a similar inversion, but is somehow more broken than src_x. (src_width-dest_height-src_y) seems to get close, but part of the source image gets obscured when you get close to its top.

I'm using driver build 4349 (waiting on the Debian package of 4363 to become available), but I expect the same problem occurs in 4363, as nothing about it was mentioned in the errata.

Hope this helps. In the mean time, I'm disabling RenderAccel...

At0mic · 06-15-03, 10:55 AM

This whole argument is stupid.
The Render extension of the driver is experimental

Hence:

Code:

(**) NVIDIA(0): Option "RenderAccel" "true"
(**) NVIDIA(0): Enabling experimental RENDER acceleration

It's also not on by default

No doubt they already know about this
Now everbody stop

ing

segflaunt · 06-15-03, 09:46 PM

Dude, the thread is really about me offering my system or my time to help test, or whatever.

I'm no master coder, but I'd like to try to offer anything I can do in case it's needed.

06-15-03, 10:55 AM	#26
At0mic Registered User Join Date: Jun 2003 Posts: 2	This whole argument is stupid. The Render extension of the driver is experimental Hence: Code: () NVIDIA(0): Option "RenderAccel" "true" () NVIDIA(0): Enabling experimental RENDER acceleration It's also not on by default No doubt they already know about this Now everbody stop ing

06-15-03, 03:01 AM	#25
amthrax \|>r4g0n Join Date: Jan 2003 Posts: 1	I'm not sure if this will help any, but I was recently debugging the Fluxbox 0.9.3 menu translucency code and discovered that the bug I was tracking was actually in nVidia's render extension. Now that I've disabled it the Fluxbox code works as expected. The problem appears to be related to the src_x and src_y parameters to XRenderComposite. The destination coordinates behave as expected (I wasn't able to test the mask coordinates from the Fluxbox code). However, the x component of the source coordinate appears to be relative to the right side of the source image, so the way to get it to look right is to use (src_width-src_x). src_y has a similar inversion, but is somehow more broken than src_x. (src_width-dest_height-src_y) seems to get close, but part of the source image gets obscured when you get close to its top. I'm using driver build 4349 (waiting on the Debian package of 4363 to become available), but I expect the same problem occurs in 4363, as nothing about it was mentioned in the errata. Hope this helps. In the mean time, I'm disabling RenderAccel...

06-15-03, 09:46 PM	#27
segflaunt Registered User Join Date: Jun 2003 Posts: 16	Dude, the thread is really about me offering my system or my time to help test, or whatever. I'm no master coder, but I'd like to try to offer anything I can do in case it's needed.